Inferring sequential structure in multivariate sparse neural signals is a very important problem for which the field needs more tools. This paper provides a principled sampling based approach with many clever tricks to tackling this significant problem efficiently and effectively.